The purpose of the study was to examine the outcome of long-term lithium treatment in consecutively admitted affective disorder patients assigned to high and low serum lithium levels. A total of 91 patients were diagnosed according to DSM-III criteria and randomly allocated to two open treatment groups in which prophylactic lithium was administered in high (serum lithium 0.8-1.0 mmol L-1) and low (serum lithium 0.5-0.8 mmol L-1) doses, respectively. The patients were followed for 2 years or until discontinuation of lithium treatment or readmission to hospital for recurrence of affective illness. The main outcome of the treatment groups was compared with Kaplan-Meier survival curves and by Cox regression analysis. A total of 31 patients (34%) completed 24 months of prophylactic lithium treatment without recurrence and readmission to hospital. In total, 18 patients (20%) suffered a recurrence on lithium, and 42 patients (46%) discontinued lithium or were lost to follow-up. No effect of treatment group was seen, either for the total patient group or for the large subgroup of bipolar patients when analysed separately. A number of patients did not maintain their original assignment to the high serum lithium levels group. The results were analysed both according to assignment and according to actual serum lithium levels. Abuse of alcohol or medication was associated with a poor outcome. Only one third of the patients completed 2 years of lithium prophylaxis successfully. No difference in the protection against recurrences was observed between patients maintained on high and low serum lithium levels.